Chest X-ray model predicted chronic kidney disease risk
The model’s risk-ranking score was 0.774 during development versus 0.713 at a second US center, where it overestimated disease risk.
5 years, independent replication, multicenter
Limit: Kidney disease was identified through diagnosis codes rather than laboratory measurements
Edited by Boris Djordjevic · AgingResearch.News
In a retrospective observational study, researchers analyzed routine chest X-rays from 97,553 US adults aged 18 or older without chronic kidney disease at baseline. They developed the model using 75,683 people at Emory University and tested it in 21,870 at the University of Illinois Chicago. Median ages were 54 and 44 years, respectively; men made up 42.1% and 34.8%.
The model used X-ray features to estimate the risk of a new chronic kidney disease diagnosis within five years, without laboratory data. Its development risk-ranking score fell to 0.737 after correction for overly favorable performance on training data. Adding clinical information gave the highest score at the second center, 0.734. Performance was strongest in people without diabetes. Median follow-up was 1,825 days at Emory and 871 days in Chicago. The authors called for recalibration and prospective evaluation before use.
Why it matters
Age was linked to new kidney disease diagnoses in the clinical model. The work asks whether existing medical images could help identify people who need further kidney assessment.
Caveats
The retrospective analysis identified disease through medical-record diagnosis codes rather than laboratory measurements. The findings do not extend to people with existing kidney disease or to inpatient or portable X-rays.
The paper
Emory University
EClinicalMedicine, Aug 4, 2026, Open access
Comments
Be the first to comment.